bde package

Subpackages

• bde.ml package
  • Submodules
  • bde.ml.loss module
    • Classes
    • Functions
    • GaussianNLLLoss
      • GaussianNLLLoss.apply_reduced()
      • GaussianNLLLoss.params
      • GaussianNLLLoss.tree_flatten()
      • GaussianNLLLoss.tree_unflatten()
    • Loss
      • Loss.apply_reduced()
      • Loss.tree_flatten()
      • Loss.tree_unflatten()
    • LossMSE
      • LossMSE.apply_reduced()
      • LossMSE.tree_flatten()
      • LossMSE.tree_unflatten()
    • NLLLoss
      • NLLLoss.apply_reduced()
      • NLLLoss.params
      • NLLLoss.tree_flatten()
      • NLLLoss.tree_unflatten()
    • flax_training_loss_wrapper_classification()
    • flax_training_loss_wrapper_regression()
  • bde.ml.models module
    • Classes
    • Functions
    • BDEEstimator
      • BDEEstimator.burn_in_loop()
      • BDEEstimator.fit()
      • BDEEstimator.get_metadata_routing()
      • BDEEstimator.get_params()
      • BDEEstimator.history_description()
      • BDEEstimator.init_inner_params()
      • BDEEstimator.load()
      • BDEEstimator.log_prior()
      • BDEEstimator.logdensity_for_batch()
      • BDEEstimator.mcmc_sampling()
      • BDEEstimator.predict()
      • BDEEstimator.predict_as_de()
      • BDEEstimator.predict_with_credibility_eti()
      • BDEEstimator.sample_from_samples()
      • BDEEstimator.save()
      • BDEEstimator.set_fit_request()
      • BDEEstimator.set_params()
      • BDEEstimator.tree_flatten()
      • BDEEstimator.tree_unflatten()
    • BasicModule
      • BasicModule.apply()
      • BasicModule.bind()
      • BasicModule.clone()
      • BasicModule.copy()
      • BasicModule.get_variable()
      • BasicModule.has_rng()
      • BasicModule.has_variable()
      • BasicModule.init()
      • BasicModule.init_with_output()
      • BasicModule.is_initializing()
      • BasicModule.is_mutable_collection()
      • BasicModule.lazy_init()
      • BasicModule.make_rng()
      • BasicModule.module_paths()
      • BasicModule.n_input_params
      • BasicModule.n_output_params
      • BasicModule.name
      • BasicModule.param()
      • BasicModule.parent
      • BasicModule.path
      • BasicModule.perturb()
      • BasicModule.put_variable()
      • BasicModule.scope
      • BasicModule.setup()
      • BasicModule.sow()
      • BasicModule.tabulate()
      • BasicModule.tree_flatten()
      • BasicModule.tree_unflatten()
      • BasicModule.unbind()
      • BasicModule.variable()
      • BasicModule.variables
    • FullyConnectedEstimator
      • FullyConnectedEstimator.fit()
      • FullyConnectedEstimator.get_metadata_routing()
      • FullyConnectedEstimator.get_params()
      • FullyConnectedEstimator.history_description()
      • FullyConnectedEstimator.init_inner_params()
      • FullyConnectedEstimator.load()
      • FullyConnectedEstimator.predict()
      • FullyConnectedEstimator.save()
      • FullyConnectedEstimator.set_params()
      • FullyConnectedEstimator.tree_flatten()
      • FullyConnectedEstimator.tree_unflatten()
    • FullyConnectedModule
      • FullyConnectedModule.apply()
      • FullyConnectedModule.bind()
      • FullyConnectedModule.clone()
      • FullyConnectedModule.copy()
      • FullyConnectedModule.do_final_activation
      • FullyConnectedModule.get_variable()
      • FullyConnectedModule.has_rng()
      • FullyConnectedModule.has_variable()
      • FullyConnectedModule.init()
      • FullyConnectedModule.init_with_output()
      • FullyConnectedModule.is_initializing()
      • FullyConnectedModule.is_mutable_collection()
      • FullyConnectedModule.layer_sizes
      • FullyConnectedModule.lazy_init()
      • FullyConnectedModule.make_rng()
      • FullyConnectedModule.module_paths()
      • FullyConnectedModule.n_input_params
      • FullyConnectedModule.n_output_params
      • FullyConnectedModule.name
      • FullyConnectedModule.param()
      • FullyConnectedModule.parent
      • FullyConnectedModule.path
      • FullyConnectedModule.perturb()
      • FullyConnectedModule.put_variable()
      • FullyConnectedModule.scope
      • FullyConnectedModule.setup()
      • FullyConnectedModule.sow()
      • FullyConnectedModule.tabulate()
      • FullyConnectedModule.tree_flatten()
      • FullyConnectedModule.tree_unflatten()
      • FullyConnectedModule.unbind()
      • FullyConnectedModule.variable()
      • FullyConnectedModule.variables
    • init_dense_model()
    • init_dense_model_jitted()
  • bde.ml.training module
    • Functions
    • jitted_evaluation_for_a_metric()
    • jitted_evaluation_over_batch()
    • jitted_training()
    • jitted_training_epoch()
    • train_step()
  • Module contents
    • Submodules
    • Example Usage
• bde.utils package
  • Submodules
  • bde.utils.configs module
    • Classes
    • General
      • General.SEED
  • Module contents
    • Modules

Submodules

Module contents

Bayesian Deep Ensembles (BDE) Package.

This package implements Bayesian Neural Networks (BNNs) with deep ensembles, combining the power of Bayesian inference with the flexibility of neural networks. The goal is to provide a robust and easy to use framework for uncertainty estimation in deep learning models.

Main Features

# TODO: Complete list - Bayesian Neural Network: Implements Bayesian inference over the parameters of a neural network, allowing for uncertainty quantification. - Integration with Popular Libraries: Compatible with major learning libraries such as sklearn. # TODO: Evaluate once implemintation is complete

Theory

# TODO: Complete

Modules

• bde.ml: Contains classes and functions related to machine learning.

• bde.utils: General utility functions and configs.

Usage

The package can be used for building and training Bayesian Neural Networks, as well as for making predictions with uncertainty estimates.

Example:

>>> # TODO: Provide an example
>>>
>>>
>>>